OtherSelf-paced; 40 to 60 hours for an existing Power Platform maker, 80+ from scratch·Training free on Microsoft Learn; associate exams are priced by country (about $165 in the US); free annual renewal

Microsoft AB-410: Building Intelligent Applications (Intelligent Applications Builder Associate)

3.9

PL-200 with an AI coat of paint, though the paint is thicker than I expected. Prompt columns, generative pages, AI Hub prompts consumed in apps and flows, and building a Copilot Studio agent from a canvas app are all in scope. Underneath, it is still a Power Platform functional consultant exam, and that is mostly a good thing.

What We Liked

Teaches AI features where they actually get used in Power Platform: inside tables, apps and flows
Cloud flows and business logic are 40 to 45%, so the core skills carry over even when the AI features change
Covers canvas app error handling, Monitor and reusable components, which PL-200 underweighted
Pairs well with AB-620 if you want to move from apps into agents

What Could Be Better

Generative pages and prompt columns are changing month to month, so some questions will date quickly
Very little on governance of AI output; responsible AI is handed off to other teams in the audience profile
AI Builder credits and licensing are the real blocker in most organisations, and the exam barely touches cost
Searches peaked at 480 in July and are already easing, which suggests most PL-200 holders are not rushing to switch

Detailed review

AB-410 replaced PL-200 when it retired on 31 July 2026. The exam is titled Building Intelligent Applications and leads to Microsoft Certified: Intelligent Applications Builder Associate. Format is the standard associate one: 120 minutes, proctored, 700 to pass, free annual renewal. Search volume for the code jumped from nothing in March to 480 in July. The outline has three parts.

Creating a foundation (25 to 30%) covers solution design with AI-enabled tools, picking built-in agents, environment and ALM strategy, and Dataverse data modelling including prompt columns and row summaries. Creating intelligent applications (25 to 30%) covers model-driven apps, generative pages built from natural language, canvas apps with components, variables, error handling and testing, and creating a Copilot Studio agent from a canvas app. Business logic and automation is the biggest block at 40 to 45%: cloud flows, connectors, approvals, prompts and models built in AI Hub and consumed in apps and flows, business rules, business process flows and calculated and formula columns. I like that the AI features are taught in context. A prompt column on a table or a prompt called from a flow is how most Power Platform teams will actually use generative AI, far more often than standalone chatbots.

The exam also asks you to evaluate when to use which kind of business logic, which is a judgement question rather than a click path, and that is where most low-code projects go wrong. My reservations are about what is missing. Nothing in the outline asks you to work out what AI Builder credits or Copilot capacity will cost, and in my experience that is what decides whether an intelligent app ships. Checking AI output quality is also absent; the audience profile hands responsible AI to governance teams. And generative pages are new enough that study material is sparse and the feature itself may look different by the time you sit the exam. Overall it is a fair, practical exam for makers and functional consultants.

If you want to go further into agents, AB-620 is the natural next step, and either one counts toward the AB-100 architect expert.

[ final ]

The verdict.

If you were going to take PL-200, take this instead; there is no choice anyway since PL-200 is retired. It is a solid low-code credential. Do not expect it to make you an AI specialist, though, because most of the marks are still for Dataverse, forms and flows.